Kaggle Competition: PUBG finish placement prediction: Top 1%

de către kamalchhirang
Kaggle Competition: PUBG finish placement prediction: Top 1%

Kaggle Competition: made a Machine Learning model to predict the PUBG finish placement prediction. I ranked at 5th place out of 1,534 teams. My prediction mean absolute error is 0.0184 and median error is 0.0533.

image of username kamalchhirang Flag of India Sikar, India

Despre mine

I would love to help you create & deploy state-of-the-art Machine Learning models into production. I have developed & deployed the following systems end-to-end: ◦ Real-time bird detection system with 99.7% accuracy used to prevent collision between birds and wind turbine using motion detection & CNN classifier. ◦ Background removal API for various objects uploaded on an e-commerce website with 100k daily visitors using UNet. ◦ Automatic product categorization API (Flask) on images uploaded on e-commerce website with 100k daily visitors with 98.9% accuracy. ◦ Real-time people & vehicle counter with 95% accuracy using EfficientDet & DeepSort deployed on AWS which is used at 300 locations. ◦ Improved sales prediction algorithm by 35% used at 100+ locations. ◦ Licence plate detection and recognition in real-time with 99.9% accuracy using EfficientDet-D0 and Gated Recurrent Convolutional Neural Network. ◦ Comment sentiment analyzer API on a website with 10k comments daily using DistilBert. ◦ Real-time face detection (99.8%+ accuracy using FaceBoxes) & face recognition (99% accuracy using ArcFace) program with age (4.1 MAE), gender (96% accuracy) & expression classifier (67% accuracy). ◦ Tic Tac Toe bot using Q-learning on 10x10 board size (total around 9.32e+156 possible combination of moves). ◦ REST API to fetching Malaysian ID card details using Image & authenticating it. ◦ AutoML for image classification, object detection & tabular data regression/classification. ◦ Personalized movie recommendation website with 300k total movies in the database, developed using Javascript, Flask (Python), SQL/MariaDB, HTML, CSS. ◦ Improved speed of machine learning inference models on CPU by 2x using OpenVino and on GPU by 3x using FP16. ◦ Identified technical system requirements based on customer’s roadmap for 70+ clients. ◦ Analyzed complex hospital patient data (MIMIC dataset) with 1000+ features and created LightGBM/CatBoost/ XGBoost models to predict mortality & hospital length of stay. ◦ Classifying pneumothorax from X-ray images using Efficientnet-B7. Skills: ◦ Proficient: Python, Javascript, SQL, PHP, HTML, CSS ◦ Familiar: C++, Matlab, Java, Hadoop ◦ Libraries: Tensorflow, Pytorch, Keras, OpenCV, LightGBM, CatBoost, XGBoost, Flask, Jupyter notebook, Numpy, ◦ Pandas, Scikit-learn, Matplotlib, Seaborn, Transformers, Scipy, Spacy. ◦ Tools: Git, AWS, Google Cloud Platform, Photoshop, Docker

USD390 USD/oră

33 păreri
5.8

Tag-uri